dsh-codebase-chat 0.19.0 → 0.22.0

package/README.md

@@ -84,19 +84,21 @@ chunks that matter, and assembles a **sourced prompt** — every answer comes ba
 ### Option A — DeepSeek Harness plugin
 
 ```bash
-pnpm add dsh-codebase-chat
-# or
-npm install dsh-codebase-chat
+dsh plugin --profile web add dsh-codebase-chat
+# without a global dsh install:
+npx -p @deepseek-ai/dsh dsh plugin --profile web add dsh-codebase-chat
 ```
 
-Then open `http://127.0.0.1:3080` and click the **Codebase Pro** button in the sidebar.
+Then restart `dsh web`, open `http://127.0.0.1:3080` and click the **Codebase Pro** button in the sidebar.
 
 ### Option B — MCP server (Cursor, Claude, Windsurf…)
 
 ```bash
-npx dsh-codebase-chat-mcp
+npx dsh-codebase-chat-mcp setup   # detects Claude/Cursor/Windsurf/VS Code and writes the config
 ```
 
+Or configure manually:
+
 ```json
 {
   "mcpServers": {
@@ -187,7 +189,7 @@ graph LR
   E --> F[Cited answer]
 ```
 
-1. **Index** — AST + regex extraction builds a symbol-level index, cached on disk and refreshed incrementally (mtime/size based — no full re-reads).
+1. **Index** — real AST extraction (Babel for JS/TS, tree-sitter WASM for Python, Go, Rust, Java, C# and PHP, regex fallback elsewhere) builds a symbol-level index, cached on disk and refreshed incrementally (mtime/size based — no full re-reads).
 2. **Retrieve** — lexical scoring plus optional local multilingual embeddings (`Xenova/paraphrase-multilingual-MiniLM-L12-v2`) select the most relevant chunks inside a token budget.
 3. **Prompt** — a language-aware prompt is assembled with module graphs, metrics, debt signals, and file excerpts.
 4. **Answer** — the model responds with `[source: file:line]` citations, `[Confidence: X%]`, and `[Severity: …]` labels.
@@ -201,6 +203,34 @@ before overwriting, enforce protected paths, and never write outside the selecte
 
 ## Configuration
 
+### `.codebase-chat.json` — per-project settings
+
+Drop this file at the project root to tune indexing, prompts and safety. Every key
+is optional; explicit CLI flags and tool arguments always take precedence, and a
+missing or malformed file falls back to built-in defaults.
+
+```json
+{
+  "lang": "en",
+  "maxTokens": 60000,
+  "ignoreDirs": ["generated", "fixtures"],
+  "ignoreFiles": ["bundle.js"],
+  "ignoreGlobs": ["src/vendor/**", "*.snap"],
+  "protectedPaths": ["src/locked", "migrations"]
+}
+```
+
+| Key | Effect |
+| --- | --- |
+| `lang` | Default prompt language (`en` or `fr`) for the CLI, MCP tools and slash commands |
+| `maxTokens` | Default context budget when the caller passes none |
+| `ignoreDirs` | Extra directory names skipped by indexing, `codebase_health` and the file tree |
+| `ignoreFiles` | Extra file names skipped the same way |
+| `ignoreGlobs` | Globs on project-relative paths — `**` spans directories, `*` one segment, `?` one char |
+| `protectedPaths` | Extra paths the apply pipeline can never patch (adds to `DSH_PROTECTED_PATHS` and built-ins) |
+
+### Environment variables
+
 | Variable | Default | Purpose |
 | --- | --- | --- |
 | `CODEBASE_CACHE_DIR` | OS cache dir | Where the index cache lives |